📝 Publications

A*-Thought: Efficient Reasoning via Bidirectional Compression for Low-Resource Settings
Xiaoang Xu, Shuo Wang, Xu Han, Zhenghao Liu, Huijia Wu, Peipei Li, Zhiyuan Liu, Maosong Sun, Zhaofeng He

• This work aims to compress long CoTs from LRMs using an A* search algorithm guided by Bidirectional Importance Scores (BIS), more details can be found on the poster.
• Code: AStar-Thought .
• Data: AStar-Thought-1k
• Models: AStar-Thought-QwQ-32B AStar-Thought-DeepSeek-R1-Distill-Qwen-32B AStar-Thought-s1.1-32B

Causality Based Front-door Defense Against Backdoor Attack on Language Models
Yiran Liu*, Xiaoang Xu*, Zhiyi Hou, Yang Yu

• This work aims to defend backdoor attack for LLMs using an algorithm named Front-door Adjustment for Backdoor Elimination (FABE) guided by Causality, more details can be found on the poster.
• Code: FABE .
• Model: FABE

HEV Generative Sandbox: A Framework for Assessing Domain-Specific Social Risks through Human-LLM Simulation
Yiran Liu, Zhiyi Hou, Xiaoang Xu, Shuo Wang, Huijia Wu, Kaicheng Yu, ChengXiang Zhai, Yang Yu

• Code: HEV-Sandbox .